A method and device for processing water bloom conditions based on ground hyperspectral images

Through the processing method based on ground hyperspectral images, effective sample points were screened, algae density sample spectral library was constructed, and Gaussian process regression model was used to solve the problem of insufficient accuracy of traditional water bloom monitoring methods, and more accurate water bloom monitoring was achieved and respond.

CN118799737BActive Publication Date: 2024-11-29YANGTZE BASIN ECOLOGY & ENVIRONMENT MONITORING & SCIENTIFIC RESEARCH CENTER YANGTZE BASIN ECOLOGY & ENVIRONMENT ADMINISTRATION MINISTRY OF ECOLOGY & ENVIRONMENT OF THE PEOPLES REPUBLIC OF CHINA +1
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202411261823.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-09-10
Publication Date
2024-11-29
Estimated Expiration
2044-09-10

Smart Images

  • Figure CN118799737B_ABST
    Figure CN118799737B_ABST
Patent Text Reader

Abstract

The present application provides a method and device for processing water bloom conditions based on ground hyperspectral images, which is used for monitoring water bloom conditions. On the basis of ground hyperspectral images, a set of practical and high-precision processing solutions is built, which helps to better understand the water bloom conditions in the basin and promote accurate response work. The method includes: based on the measured chlorophyll a content data, determining the target geographical location data of the effective sample point, then determining the target spectral information of the target ground hyperspectral image, and matching the target spectral information with the measured algae density data to obtain an algae density sample spectral library; based on the algae density sample spectral library, constructing a corresponding Gaussian process regression model; based on the Gaussian process regression model, determining the three bands that achieve the best algae density response capability; calculating the algae density index based on the three bands, and determining whether there is a water bloom; processing the geographical distribution map of the water bloom algae density index, and classifying it.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of water bloom monitoring, and in particular to a method and device for processing water bloom conditions based on ground hyperspectral images. Background Art

[0002] Algal bloom is a natural ecological phenomenon of algae mass reproduction in freshwater bodies. It is mainly caused by the large amount of nitrogen and phosphorus wastewater entering the water body from daily life and industrial and agricultural production, which leads to the mass reproduction of blue algae, green algae and diatoms, making the water body appear blue or green. Simply put, algal bloom is seen by the naked eye as algae aggregation or algae particles suspended on the surface of the water body. Algal bloom has its adverse effects, such as: increasing the cost of drinking water treatment; algal bloom has various hazards, releasing toxins into the air, endangering human health; consuming underwater oxygen, endangering aquaculture; increasing greenhouse gas emissions, endangering ecological balance.

[0003] Generally speaking, the intensity of algal blooms is determined by the density of algae in the water. However, traditional methods of monitoring algae density often monitor the density of algae in water samples through manual laboratory testing or rapid testing with portable instruments, or monitor changes in chlorophyll a through satellite remote sensing to measure the degree of eutrophication of the water body, thereby indirectly demonstrating the intensity of algal blooms.

[0004] However, the inventors of the present application have found that most studies roughly calculate the correlation between algae density and chlorophyll a content through linear regression and correlation analysis methods to indirectly reflect the intensity of water blooms, or directly establish a regression relationship between spectral data and algae density to obtain the inversion result of algae density. However, the obtained water bloom conditions reflect the problem of insufficient stability in accuracy in practice, and in most cases the accuracy is low. Summary of the invention

[0005] The present application provides a method and device for processing water bloom conditions based on ground hyperspectral images, which are used for monitoring water bloom conditions. On the basis of ground hyperspectral images, a set of practical and high-precision processing solutions are built, which helps to better understand the water bloom conditions in the basin and promote accurate response work.

[0006] In a first aspect, the present application provides a method for processing water bloom conditions based on ground hyperspectral images, the method comprising:

[0007] For the target watershed, the corresponding initial ground hyperspectral image, the measured algae density data of the ground measurement points, the measured chlorophyll a content data of the ground measurement points, and the initial geographical location data of the ground measurement points are obtained respectively;

[0008] Based on the measured chlorophyll a content data, the effective sample points are screened out, the target geographical location data of the effective sample points are determined from the initial geographical location data, and then the target spectral information of the target ground hyperspectral image corresponding to the effective sample points is determined from the initial ground hyperspectral image according to the target geographical location data, and the target spectral information is matched with the measured algae density data to obtain the algae density sample spectral library;

[0009] Based on the algae density sample spectral library, the corresponding Gaussian process regression model is constructed, where the input of the Gaussian process regression model is spectral data and the output is algae density data.

[0010] The algae density response capabilities of different spectral sensor bands were determined based on the Gaussian process regression model, and the three bands with the best algae density response capabilities were determined;

[0011] The algae density index is calculated based on the three bands, and the presence of algae bloom is determined by combining the corresponding algae density index threshold;

[0012] The geographical distribution maps of algal bloom density indices of different watersheds including the target watershed are processed and classified according to the size of the algal density index.

[0013] In a second aspect, the present application provides a device for processing water bloom conditions based on ground hyperspectral images, the device comprising:

[0014] The initial data acquisition unit is used to acquire the corresponding initial ground hyperspectral image, the measured algae density data of the ground measured points, the measured chlorophyll a content data of the ground measured points, and the initial geographical location data of the ground measured points for the target watershed;

[0015] A spectral library construction unit is used to screen out valid sample points based on the measured chlorophyll a content data, determine the target geographical location data of the valid sample points from the initial geographical location data, and then determine the target spectral information of the target ground hyperspectral image corresponding to the valid sample points from the initial ground hyperspectral image according to the target geographical location data, and match the target spectral information with the measured algae density data to obtain an algae density sample spectral library;

[0016] A model building unit is used to build a corresponding Gaussian process regression model based on the algae density sample spectrum library, wherein the input of the Gaussian process regression model is the spectrum data and the output is the algae density data;

[0017] A band determination unit, used for determining the algae density response capabilities of different spectral sensor bands based on a Gaussian process regression model, and determining three bands with the best algae density response capabilities;

[0018] An index calculation unit, used to calculate the algae density index based on the three bands, and determine whether there is an algae bloom in combination with a corresponding algae density index threshold;

[0019] The distribution map processing unit is used to process the geographical distribution map of the water bloom algae density index of different basins including the target basin, and classify them according to the size of the algae density index.

[0020] From the above content, it can be concluded that the present application has the following beneficial effects:

[0021] For the monitoring target of algal bloom, this application obtains the corresponding initial ground hyperspectral image, the measured algae density data of the ground measured points, the measured chlorophyll a content data of the ground measured points, and the initial geographical location data of the ground measured points for the target watershed, and then screens out the valid sample points based on the measured chlorophyll a content data, determines the target geographical location data of the valid sample points from the initial geographical location data, and then determines the target spectral information of the target ground hyperspectral image corresponding to the valid sample points from the initial ground hyperspectral image based on the target geographical location data, and matches the target spectral information with the measured algae density data to obtain the algae density sample spectral library, and then constructs the corresponding Gaussian process based on the algae density sample spectral library. Regression model, in which the input of the Gaussian process regression model is spectral data, and the output is algae density data. At this time, the algae density response capacity of different spectral sensor bands is continued to be determined based on the Gaussian process regression model, and the three bands with the best algae density response capacity are determined. Then the algae density index is calculated based on the three bands, and the corresponding algae density index threshold is combined to determine whether there is an algal bloom. And so on, the geographical distribution map of the algae density index of water blooms in different basins including the target basin is finally processed, and classified according to the size of the algae density index. Therefore, based on the ground hyperspectral imagery, a set of practical and high-precision processing solutions is built, which helps to better understand the water bloom situation in the basin and promote accurate response work. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0023] Figure 1 A schematic diagram of a process of the water bloom processing method based on ground hyperspectral images in this application;

[0024] Figure 2 A schematic diagram of a scenario for selecting effective sample points for this application;

[0025] Figure 3 A schematic diagram of a scenario of the existing algal bloom degree classification mechanism;

[0026] Figure 4 A schematic diagram of a scenario for verifying the algae bloom identification results of this application;

[0027] Figure 5 A schematic diagram of the structure of a device for processing water bloom conditions based on ground hyperspectral images in this application. DETAILED DESCRIPTION

[0028] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of this application.

[0029] The terms "first", "second", etc. in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments described here can be implemented in an order other than that illustrated or described here. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or modules is not necessarily limited to those steps or modules clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products or devices. The naming or numbering of steps in this application does not mean that the steps in the method flow must be executed in the time / logical sequence indicated by the naming or numbering. The process steps that have been named or numbered can change the execution order according to the technical purpose to be achieved, as long as the same or similar technical effects can be achieved.

[0030] The division of modules in this application is a logical division. There may be other division methods when it is implemented in actual applications. For example, multiple modules can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, and the indirect coupling or communication connection between modules can be electrical or other similar forms, which are not limited in this application. In addition, the modules or submodules described as separate components may or may not be physically separated, may or may not be physical modules, or may be distributed in multiple circuit modules, and some or all of the modules may be selected according to actual needs to achieve the purpose of the present application.

[0031] The present application provides a method and device for processing water bloom conditions based on ground hyperspectral images, which are used for monitoring water bloom conditions. Based on ground hyperspectral images, a set of practical and high-precision processing solutions are established, which helps to better understand the water bloom conditions in the basin and promote accurate response work.

[0032] Next, the water bloom processing method based on ground hyperspectral imagery provided by the present application is introduced.

[0033] First, see Figure 1 , Figure 1 A schematic flow chart of the water bloom processing method based on ground hyperspectral images provided by the present application is shown. The water bloom processing method based on ground hyperspectral images provided by the present application may specifically include the following steps S101 to S106:

[0034] Step S101, for the target watershed, respectively obtain the corresponding initial ground hyperspectral image, the measured algae density data of the ground measured points, the measured chlorophyll a content data of the ground measured points, and the initial geographical location data of the ground measured points;

[0035] It can be understood that water phenomena occur in river basins (can also be referred to as water areas). When there is a need to deal with water blooms in a certain river basin, high-precision water blooms can be handled through this application solution, which helps to determine response work such as maintaining observation or timely handling.

[0036] Among them, when carrying out processing, the present application is in the form of processing local basins in sequence or processing multiple local basins at the same time to finally obtain the water bloom situation of the entire basin. The treatment of the local basins is the same, so from the perspective of the local basin, it is recorded as the target basin to expand the description of the present application scheme.

[0037] It is easy to see that the initial data involved in the present application scheme include initial ground hyperspectral images (a type of satellite image collected by a satellite) corresponding to the target watershed, measured algae density data of ground measured points, measured chlorophyll a content data of ground measured points, and initial geographical location data of ground measured points.

[0038] It is understandable that these four types of data are mature concepts in themselves, and the way to obtain the data is relatively simple. Existing technologies can be directly used. For example, the measured algae density data, measured chlorophyll a content data and initial geographic location data can be obtained through manual laboratory testing or rapid testing with portable instruments. No further explanation will be given here.

[0039] After obtaining these four types of data, it is easy to understand that corresponding preprocessing operations can also be involved to enhance the data quality.

[0040] As an exemplary embodiment, the method of the present application may further include:

[0041] The initial ground hyperspectral image is preprocessed, including radiation calibration, geometric correction, image mosaicking, spatial cropping, atmospheric correction, cloud removal, and band resampling.

[0042] It is understandable that the several preprocessing means involved here can adopt existing technologies themselves, and the focus of the embodiment here is to point out that these specific preprocessing means can be used to achieve high-quality preprocessing effects on the initial ground hyperspectral image.

[0043] Among them, there may be corresponding general technical specifications for radiation calibration, geometric correction, spatial clipping and atmospheric correction. Therefore, these preprocessing methods can be understood as meeting the corresponding general technical specifications during preprocessing, or can be understood as directly processing the source data to a state that meets the corresponding general technical specifications. For example, geometric correction may involve the requirements of chapter 4.6.2 of Technical Specifications for Remote Sensing and Ground Monitoring and Evaluation of Water Bloom (Trial) HJ 1098-2020, spatial clipping may involve the requirements of chapter 4.6.3 of Technical Specifications for Remote Sensing and Ground Monitoring and Evaluation of Water Bloom (Trial) HJ 1098-2020, radiation calibration may involve the requirements of chapter 4.6.4 of Technical Specifications for Remote Sensing and Ground Monitoring and Evaluation of Water Bloom (Trial) HJ 1098-2020, and atmospheric correction may involve the requirements of chapter 4.6.4 of Technical Specifications for Remote Sensing and Ground Monitoring and Evaluation of Water Bloom (Trial) HJ 1098-2020. Declouding is to remove the clouds included in the satellite remote sensing process. As an example, it can be achieved by using the Fast Line of Sight atmospheric analysis of spectral hypercubes (FLAASH) algorithm.

[0044] Band resampling can be understood as sampling the source data into data of corresponding bands through sampling means. As an example, the bands of ground hyperspectral image data can be sampled to 443nm aerosol band, 490nm blue band, 560nm green band, 665nm red band, 705nm red band, 740nm red band, 783nm red band, 843nm red band and 865nm red band.

[0045] Among them, it can be seen that different specific bands are involved in the red band. In actual circumstances, the red band can also be divided into visible light band, near-infrared band and other types. Considering that several specific band values ​​have been listed here, they are collectively referred to as red bands, and the specific band values ​​are used to distinguish them.

[0046] Similarly, the measured algae density data may involve the requirements of Chapter 5 (HJ 1098-2020) of the Technical Specifications for Remote Sensing and Ground Monitoring and Evaluation of Algal Blooms (Trial), and the measured chlorophyll a content data may also involve the requirements of Chapter 5 (HJ 1098-2020) of the Technical Specifications for Remote Sensing and Ground Monitoring and Evaluation of Algal Blooms (Trial).

[0047] In addition, it can be understood that the four types of initial data obtained in step S101 here may also be data obtained after performing corresponding preprocessing on the source data.

[0048] Step S102, based on the measured chlorophyll a content data of the ground measured points, the valid sample points are screened out, the target geographical location data of the valid sample points are determined from the initial geographical location data, and then the target spectral information of the target ground hyperspectral image corresponding to the valid sample points is determined from the initial ground hyperspectral image according to the target geographical location data, and the target spectral information is matched with the measured algae density data to obtain an algae density sample spectral library;

[0049] After obtaining the above four types of initial data, corresponding processing can be carried out based on the measured chlorophyll a content data to finally obtain the algae density sample spectral library corresponding to the target watershed required by the present application.

[0050] Specifically, it can be understood that according to data statistical requirements, corresponding filtering means can be configured to screen out data points that meet the requirements from the data points involved in the measured chlorophyll a content data as valid data points involved in the subsequent data processing process of this application scheme.

[0051] As an exemplary embodiment, based on the measured chlorophyll a content data of the ground measured points, effective sample points are screened out, which may specifically include:

[0052] The measured chlorophyll a content data are equally divided into multiple groups of chlorophyll a content data, and each group of chlorophyll a content data intervals is cycled to screen out the data points of chlorophyll a content and corresponding algae density data in each specified interval, and the data points in the specified interval with algae density data values ​​in the range of 5% to 95% are taken as valid sample points.

[0053] Among them, it can be understood that the interval (data interval) involved here is formed in the process of dividing the measured chlorophyll a content data. In actual situations, it can be specified by the user or equipment. In this way, combined with the judgment condition of which data points are within the range of 5% to 95% of the algae density data value, valid sample points are screened out. Conversely, data points with algae density data values ​​within the range of 5% to 95% are invalid sample points.

[0054] As an example, see Figure 2 A schematic diagram of a scenario for screening effective sample points in the present application is shown for a more vivid understanding.

[0055] In this way, after the valid sample points are screened out, the corresponding target first position data, ie, the geographic location data of the valid sample points, can be extracted from the initial geographic location data obtained previously based on the valid sample points.

[0056] Next, at the location indicated by the target geographic location data, the target ground hyperspectral image corresponding to the valid sample point is determined from another initial ground hyperspectral image, and the target spectral information corresponding to the target ground hyperspectral image is extracted for use in building an algae density sample spectral library corresponding to the target watershed.

[0057] As an example, the spectral information of the ground hyperspectral image corresponding to the effective sample points can be extracted by using ArcGIS 10.8 software.

[0058] In addition, it is worth adding that the above processing is based on the same time dimension. For example, the four initial data obtained in step S101 are in the same time dimension. Therefore, there is no need to align the time granularity during the processing, because the data itself is in the same time granularity by default. If the four initial data obtained in step S101 are not necessarily in the same time dimension or not all in the same time dimension, they need to be aligned in the time dimension. For example, the spectral information of the ground hyperspectral image corresponding to the time of the effective sample point can be extracted here.

[0059] At this point, the target spectral information can be matched / paired with the measured algae density data to obtain an algae density sample spectral library.

[0060] It is easy to understand from the literal meaning that the algae density sample spectral library is a database that identifies the pairing relationship between spectral information and algae density for effective sample points.

[0061] Among them, for the algae density sample spectral library, as an exemplary embodiment, the different spectral information involved corresponds to the exemplary description content mentioned in the previous band resampling, and may involve 9 bands, specifically including 443nm aerosol band, 490nm blue band, 560nm green band, 665nm red band, 705nm red band, 740nm red band, 783nm red band, 843nm red band and 865nm red band.

[0062] Step S103, based on the algae density sample spectrum library, construct a corresponding Gaussian process regression model, wherein the input of the Gaussian process regression model is the spectrum data, and the output is the algae density data;

[0063] After obtaining the algae density sample spectral library corresponding to the target watershed, a Gaussian process regression model can be constructed based on the database of the pairing relationship between the spectral information recorded for the effective sample points and the algae density.

[0064] Among them, the Gaussian Process Regression model (GPR) can also be referred to as the GPR model.

[0065] It can be understood that this application, based on a limited amount of algae density sample spectral library, configures a Gaussian process regression model to delicately process the algae density data corresponding to the spectral data, which can lay a good foundation for the algae density response capabilities of different spectral sensor bands required subsequently.

[0066] As an exemplary embodiment, based on the algae density sample spectrum library, a corresponding Gaussian process regression model is constructed, which may specifically include:

[0067] 1) Select the initialized Gaussian process regression model, take the spectral information as the independent variable, take the algae density data as the dependent variable, select the Gaussian process kernel function and leave-one-out method as iterative optimization parameters, establish the Gaussian process regression model, and record it as f(x);

[0068] 2) Using the Monte Carlo method to simulate N sample points of d-type spectral information, an N×2d spectral information sample matrix is ​​generated, where each row of the N×2d spectral information sample matrix is ​​a sampling point in the 2-dimensional hyperspace (this is related to the probability distribution of the spectral information as input);

[0069] 3) Use the first d columns of the N×2d spectral information sample matrix as the spectral information matrix A and the remaining d columns as the spectral information matrix B to (effectively) give two independent samples of spectral information for N points in the d-dimensional unit hypercube;

[0070] 4) Construct dN×d spectral information matrix , i=1,2,...,d, dN×d spectral information matrix The i-th column of is the i-th column of the spectral information matrix B, and the remaining columns are the spectral information matrix A;

[0071] 5) Spectral information matrix A, spectral information matrix B and dN×d spectral information matrix N(d+2) points in the input space are in the spectral information matrix A, spectral information matrix B and dN×d spectral information matrix Run the (Gaussian process regression model) f(x) for each design point in , and give a total of N(d+2) values ​​of f(A), f(B), and f( )value.

[0072] It can be seen that here is how to build a corresponding Gaussian process regression model based on the algae density sample spectral library in actual application, and use Monte Carlo simulation to generate spectral information sample points, and bring them into the model to obtain the contribution of spectral information to algae density data (that is, the response degree of each spectral information to algae density). Among them, the higher the contribution of spectral information, the more relevant the spectral information is to the algae density data. Therefore, the design here is to determine the spectral information with high or even the highest correlation for the construction of subsequent indexes, thereby providing a set of specific implementation supporting solutions to effectively improve the construction accuracy and efficiency of the subsequent algae density index.

[0073] Step S104, determining the algae density response capabilities of different spectral sensor bands based on the Gaussian process regression model, and determining three bands with the best algae density response capabilities;

[0074] As mentioned earlier, the configured Gaussian process regression model is used to determine the algae density response capabilities of different spectral sensor bands. Specifically, when the satellite collects ground hyperspectral images, different usage conditions / situations will be involved, which will affect the differences in the spectral information behind them. This difference specifically involves differences in sensor design, satellite shooting locations, angles, weather, and other aspects.

[0075] Therefore, the inventors of the present application believe that, in order to invert the algae density, one can try to analyze the response degree of the spectral information captured by the sensor to the algae density based on the specific spectral sensor.

[0076] Under this design direction, based on the characteristic screening method of remote sensing spectral response, the global sensitivity analysis method can be used to quantify the contribution of spectral parameters of different sensors to algae density parameters. Then, based on the relevant results of the calculated spectral response index of each band of the sensor, the best band for developing the inverted algae density parameters is selected, and the three bands with the best algae density response capability are preferably used as the input for the algae density index processing specially designed for the present application.

[0077] As an exemplary embodiment, determining the algae density response capability of different spectral sensor bands based on the Gaussian process regression model may specifically include:

[0078] The spectral response index is calculated using the following formula:

[0079] E X~i (Var Xi (Y|X ~i ))≈ ,

[0080] S Ti =E X~i (Var Xi (Y|X ~i )) / Var(Y),

[0081] Among them, E X~i (Var Xi (Y|X ~i )) means when X takes X ~i The conditional variance of the output when Y is represented by X ~i The algae density data obtained, X ~i refers to the i-th column of the input spectral information matrix, j represents the current sample point among N sample points, Var(Y) is the variance of algae density data, S Ti is the spectral response index of algae density data, which is proportional to the responsiveness of algae density.

[0082] It is easy to see that the embodiment here is configured on the basis of the Gaussian process regression model configured in the previous embodiment, specifically involving the f(A) value and f( ) values, thus providing a specific implementation plan for quantifying the algae density response capability, and can effectively improve the processing efficiency and effectiveness of constructing the algae density index.

[0083] Step S105, calculating the algae density index based on the three bands, and determining whether there is an algae bloom in combination with the corresponding algae density index threshold;

[0084] After obtaining the three bands with the best algae density response capability, they can be substituted into the algae density index quantification formula configured in this application to well reflect the algae bloom situation in the target watershed.

[0085] Among them, it should be noted that the algae density index involved in this application is not the algae density, but a new indicator constructed by this application considering how to reflect the algae density situation / water bloom degree.

[0086] In this way, the calculated algae density index is compared with the algae density index threshold. Obviously, if it is greater than, it can be considered that there is an algae bloom, otherwise it can be considered that there is no algae bloom.

[0087] As an exemplary embodiment, the algae density index is calculated based on three bands, which may specifically include:

[0088] The enumeration method is used to select two bands from the three bands and substitute them into the construction process of the algae density index quantitative formula. In each construction process of the algae density index quantitative formula, the two currently selected bands ρ1 and ρ2 are substituted into the quantitative formulas of the ratio vegetation index VI1, the normalized vegetation index VI2, and the difference vegetation index VI3 for calculation. Then, the combination formulas composed of the three elements of the ratio vegetation index VI1, the normalized vegetation index VI2 and the difference vegetation index VI3 are exhaustively enumerated through the constraint method involving only addition, subtraction, multiplication and division operations. After obtaining different combination formulas obtained by different combinations of the two bands selected from the three bands, the degree of fit between the algae density index calculated by different combination formulas and the measured algae density data is evaluated (the closer the fit, the more accurately the algae density and water bloom can be reflected), and the combination formula with the best fit is confirmed as the algae density index quantitative formula suitable for the target basin, and the algae density index value calculated under the algae density index quantitative formula suitable for the target basin is determined, among which:

[0089] VI1=ρ1 / ρ2,

[0090] VI2=(ρ1-ρ2) / (ρ1+ρ2),

[0091] VI3=ρ1-ρ2,

[0092] Among them, if the algae density index is greater than the threshold, there is an algal bloom, and if the algae density index is less than the threshold, there is no algal bloom.

[0093] It should be understood that in the process of constructing the above-mentioned algae density index quantification formula, for different river basins or river basins under different time conditions, the algae density index quantification formula finally determined is often different in expression. Therefore, when constructing the algae density index quantification formula for the current target river basin, it has the characteristic of high adaptability, so that the one-to-one matching algae density index quantification formula can be used to accurately quantify the degree of water bloom in the current target river basin.

[0094] It can be seen that in the embodiment here, a specific mathematical expression is constructed for how to calculate the algae density index introduced in this application. This mathematical expression is not easily affected by changes in environmental observation conditions such as aerosol type, aerosol thickness, solar altitude angle, flares, etc., and can penetrate thin clouds more effectively. The threshold for extracting water blooms is more stable. Similar to the previous embodiment, a set of specific implementation supporting solutions is provided, which has the characteristics of high precision and can accurately reflect the degree of water bloom in the target basin, thus having better practical significance.

[0095] Step S106, processing the geographical distribution map of water bloom algae density index of different watersheds including the target watershed, and classifying them according to the size of the algae density index.

[0096] It can be understood that the above steps S101 to S105 determine the algae density index and water bloom situation of the target basin. For the entire basin to be monitored, the processing of other local basins can also be involved, and the corresponding algae density index and water bloom situation can be determined in the same way.

[0097] Finally, the algae density index and water bloom conditions of all local watersheds can be integrated to obtain the geographical distribution map of water bloom algae density index in different watersheds including the target watershed, which will help to present the algae density index and water bloom conditions of the entire watershed in a more intuitive visual way.

[0098] Correspondingly, the method of the present application may also include display processing of the geographical distribution map of the water bloom algae density index obtained here.

[0099] Furthermore, since the size of the algae density index is proportional to the severity of the algae bloom, it can also be classified according to the size of the algae density index, and reflect the severity of the algae density index and / or algae bloom in different watersheds, and can also be distinguished more finely through different display colors.

[0100] For the classification of algal bloom, relevant specifications may be involved in specific operations. For example, the algal bloom classification standards given in Section 6.3 of the Technical Specifications for Algal Bloom Remote Sensing and Ground Monitoring and Evaluation (Trial) HJ 1098-2020 can be referred to. Figure 3A scenario diagram of the existing algal bloom degree classification mechanism is shown for a more vivid understanding.

[0101] It is important to understand that Figure 3 The classification content shown is configured based on algae density. In this application scheme, the classification of water bloom degree is based on the algae density index, which can be referred to Figure 2 The grading mechanism shown in the prior art specification is used to configure algae density indexes of different magnitudes to achieve an adaptive grading effect.

[0102] In the specific research and development process, the inventor of this application input 93 samples for verification. The results showed that 18 were misclassified as water blooms, 75 were correctly identified, and the overall recognition accuracy was 80.65%. Figure 4 A schematic diagram of a scenario for verifying the water bloom identification results of the present application shows intuitively that the present application scheme has a better performance in terms of prediction accuracy, and thus has a better application value.

[0103] In general, for the monitoring target of algal bloom, this application obtains the corresponding initial ground hyperspectral images, the measured algae density data of the ground measured points, the measured chlorophyll a content data of the ground measured points, and the initial geographical location data of the ground measured points for the target watershed, and then screens out the valid sample points based on the measured chlorophyll a content data, determines the target geographical location data of the valid sample points from the initial geographical location data, and then determines the target spectral information of the target ground hyperspectral image corresponding to the valid sample points from the initial ground hyperspectral image based on the target geographical location data, and matches the target spectral information with the measured algae density data to obtain the algae density sample spectral library, and then constructs the corresponding high A Gaussian process regression model is proposed, in which the input of the model is spectral data and the output is algae density data. At this time, the algae density response capabilities of different spectral sensor bands are continued to be determined based on the Gaussian process regression model, and the three bands with the best algae density response capabilities are determined. Then the algae density index is calculated based on the three bands, and the corresponding algae density index threshold is used to determine whether there is an algal bloom. By analogy, the geographical distribution map of the algae density index of water blooms in different basins including the target basin is finally processed, and classified according to the size of the algae density index. Therefore, a practical and high-precision processing solution is established based on the ground hyperspectral imagery, which helps to better understand the water bloom situation in the basin and promote accurate response work.

[0104] The above is an introduction to the water bloom situation processing method based on ground hyperspectral images provided in this application. In order to facilitate better implementation of the water bloom situation processing method based on ground hyperspectral images provided in this application, this application also provides a water bloom situation processing device based on ground hyperspectral images from the perspective of functional modules.

[0105] See also Figure 5 , Figure 5 This is a schematic diagram of a structure of a water bloom processing device based on ground hyperspectral images in this application. In this application, the water bloom processing device 500 based on ground hyperspectral images may specifically include the following structure:

[0106] The initial data acquisition unit 501 is used to acquire the corresponding initial ground hyperspectral image, the measured algae density data of the ground measured points, the measured chlorophyll a content data of the ground measured points, and the initial geographical location data of the ground measured points for the target watershed;

[0107] The spectral library construction unit 502 is used to screen out valid sample points based on the measured chlorophyll a content data, determine the target geographical location data of the valid sample points from the initial geographical location data, and then determine the target spectral information of the target ground hyperspectral image corresponding to the valid sample points from the initial ground hyperspectral image according to the target geographical location data, and match the target spectral information with the measured algae density data to obtain an algae density sample spectral library;

[0108] The model building unit 503 is used to build a corresponding Gaussian process regression model based on the algae density sample spectrum library, wherein the input of the Gaussian process regression model is the spectrum data and the output is the algae density data;

[0109] The band determination unit 504 is used to determine the algae density response capabilities of different spectral sensor bands based on the Gaussian process regression model, and determine three bands with the best algae density response capabilities;

[0110] An index calculation unit 505 is used to calculate the algae density index based on the three bands, and determine whether there is an algae bloom in combination with a corresponding algae density index threshold;

[0111] The distribution map processing unit 506 is used to process the geographical distribution maps of the water bloom algae density index of different watersheds including the target watershed, and classify them according to the size of the algae density index.

[0112] In an exemplary embodiment, the initial data acquisition unit 501 is further configured to:

[0113] The initial ground hyperspectral image is preprocessed, including radiation calibration, geometric correction, image mosaicking, spatial clipping, atmospheric correction, cloud removal and band resampling.

[0114] In another exemplary embodiment, the spectral library construction unit 502 is specifically configured to:

[0115] The measured chlorophyll a content data are equally divided into multiple groups of chlorophyll a content data, and each group of chlorophyll a content data intervals is cycled to screen out the data points of chlorophyll a content and corresponding algae density data in each specified interval, and the data points in the specified interval with algae density data values ​​in the range of 5% to 95% are taken as valid sample points.

[0116] In another exemplary embodiment, the algae density sample spectral library involves 9 bands, including an aerosol band of 443nm, a blue band of 490nm, a green band of 560nm, a red band of 665nm, a red band of 705nm, a red band of 740nm, a red band of 783nm, a red band of 843nm and a red band of 865nm.

[0117] In another exemplary embodiment, the model building unit 503 is specifically configured to:

[0118] Select the initialized Gaussian process regression model, take the spectral information as the independent variable, take the algae density data as the dependent variable, select the Gaussian process kernel function and leave-one-out method as iterative optimization parameters, establish the Gaussian process regression model, and record it as f(x);

[0119] Monte Carlo method is used to simulate N sample points of d-type spectral information to generate an N×2d spectral information sample matrix, where each row of the N×2d spectral information sample matrix is ​​a sampling point in the 2-dimensional hyperspace.

[0120] Using the first d columns of the N×2d spectral information sample matrix as the spectral information matrix A and the remaining d columns as the spectral information matrix B to give two independent samples of spectral information for N points in the d-dimensional unit hypercube;

[0121] Constructing a dN×d spectral information matrix , i=1,2,...,d, dN×d spectral information matrix The i-th column of is the i-th column of the spectral information matrix B, and the remaining columns are the spectral information matrix A;

[0122] The spectral information matrix A, spectral information matrix B and dN×d spectral information matrix N(d+2) points in the input space are in the spectral information matrix A, spectral information matrix B and dN×d spectral information matrix Run f(x) for each design point in , and give a total of N(d+2) values ​​of f(A), f(B), and f( )value.

[0123] In another exemplary embodiment, the band determining unit 504 is specifically configured to:

[0124] The spectral response index is calculated using the following formula:

[0125] E X~i (Var Xi (Y|X ~i ))≈ ,

[0126] S Ti =E X~i (Var Xi (Y|X ~i )) / Var(Y),

[0127] Among them, E X~i (Var Xi (Y|X ~i )) means when X takes X ~i The conditional variance of the output when Y is represented by X ~i The algae density data obtained, X ~i refers to the i-th column of the input spectral information matrix, j represents the current sample point among N sample points, Var(Y) is the variance of algae density data, S Ti is the spectral response index of algae density data, which is proportional to the responsiveness of algae density.

[0128] In another exemplary embodiment, the index calculation unit 505 is specifically configured to:

[0129] The enumeration method is used to select two bands from the three bands and substitute them into the construction process of the algae density index quantitative formula. In each construction process of the algae density index quantitative formula, the two currently selected bands ρ1 and ρ2 are substituted into the quantitative formulas of the ratio vegetation index VI1, the normalized vegetation index VI2, and the difference vegetation index VI3 for calculation. Then, the combination formula consisting of the three elements of the ratio vegetation index VI1, the normalized vegetation index VI2 and the difference vegetation index VI3 is exhaustively enumerated through the constraint method involving only addition, subtraction, multiplication and division operations. After obtaining different combination formulas obtained by different combinations of the two bands selected from the three bands, the degree of fit between the algae density index calculated by different combination formulas and the measured algae density data is evaluated, and the combination formula with the best fit is confirmed as the algae density index quantitative formula adapted to the target basin. At the same time, the algae density index value calculated under the algae density index quantitative formula adapted to the target basin is determined, wherein:

[0130] VI1=ρ1 / ρ2,

[0131] VI2=(ρ1-ρ2) / (ρ1+ρ2),

[0132] VI3=ρ1-ρ2,

[0133] Among them, if the algae density index is greater than the threshold, there is an algal bloom, and if the algae density index is less than the threshold, there is no algal bloom.

[0134] Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working process of the above-described water bloom processing device based on ground hyperspectral imagery and its corresponding units can be referred to as follows: Figure 1 The description of the method for processing water bloom based on ground hyperspectral images in the corresponding embodiment will not be repeated here in detail.

[0135] The above is a detailed introduction to the water bloom processing method and device based on ground hyperspectral images provided by the present application. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea; at the same time, for technical personnel in this field, according to the idea of ​​the present application, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A method for processing water bloom conditions based on ground hyperspectral images, characterized in that: The method comprises: For the target watershed, respectively obtain the corresponding initial ground hyperspectral image, the measured algae density data of the ground measured points, the measured chlorophyll a content data of the ground measured points, and the initial geographical location data of the ground measured points; Based on the measured chlorophyll a content data, valid sample points are screened out, target geographic location data of the valid sample points are determined from the initial geographic location data, and then target spectral information of the target ground hyperspectral image corresponding to the valid sample points is determined from the initial ground hyperspectral image according to the target geographic location data, and the target spectral information is matched with the measured algae density data to obtain an algae density sample spectral library; Based on the algae density sample spectral library, a corresponding Gaussian process regression model is constructed, wherein the input of the Gaussian process regression model is spectral data, and the output is algae density data; Determine the algae density response capabilities of different spectral sensor bands based on the Gaussian process regression model, and determine three bands with the best algae density response capabilities; Calculating the algae density index based on the three bands, and determining whether there is an algae bloom in combination with a corresponding algae density index threshold; Processing geographical distribution maps of water bloom algae density indices of different watersheds including the target watershed, and classifying them according to the size of the algae density index; Based on the algae density sample spectral library, a corresponding Gaussian process regression model is constructed, including: Selecting an initialized Gaussian process regression model, taking the spectral information as an independent variable, taking the algae density data as a dependent variable, selecting a Gaussian process kernel function and a leave-one-out method as iterative optimization parameters, establishing the Gaussian process regression model, and recording it as f(x); Using the Monte Carlo method to simulate N sample points of d-type spectral information, an N×2d spectral information sample matrix is ​​generated, wherein each row of the N×2d spectral information sample matrix is ​​a sampling point in a 2-dimensional hyperspace; Using the first d columns of the N×2d spectral information sample matrix as spectral information matrix A and the remaining d columns as spectral information matrix B to give two independent samples of spectral information for N points in a d-dimensional unit hypercube; Constructing a dN×d spectral information matrix i=1,2,...,d, the dN×d spectral information matrix The i-th column of is the i-th column of the spectral information matrix B, and the remaining columns are the spectral information matrix A; The spectral information matrix A, the spectral information matrix B and the dN×d spectral information matrix N(d+2) points in the input space are in the spectral information matrix A, the spectral information matrix B and the dN×d spectral information matrix Run f(x) for each design point in , and give a total of N(d+2) values ​​of f(A), f(B), and value; The method of determining the algae density response capability of different spectral sensor bands based on the Gaussian process regression model includes: The spectral response index is calculated by the following formula: S Ti =E X~i (There is Xi (Y|X ~i )) / Yes(Y), Among them, E X~i (Var Xi (Y|X ~i )) means when X takes X ~i The conditional variance of the output when Y is represented by X ~i The algae density data obtained, X ~i refers to the i-th column of the input spectral information matrix, j represents the current sample point among the N sample points, Var(Y) is the variance of the algae density data, S Ti is a spectral response index of algae density data, wherein the spectral response index is proportional to the responsiveness of the algae density; The algae density index is calculated based on the three bands, including: The enumeration method is adopted to select two bands from the three bands and substitute them into the construction process of the algae density index quantification formula, wherein, in each construction process of the algae density index quantification formula, the two currently selected bands ρ1 and ρ2 are substituted into the quantification formulas of the ratio vegetation index VI1, the normalized vegetation index VI2, and the difference vegetation index VI3 for calculation, and then the combination formula consisting of the three elements of the ratio vegetation index VI1, the normalized vegetation index VI2 and the difference vegetation index VI3 is exhaustively enumerated in a constraint mode involving only addition, subtraction, multiplication and division operations, and different combination formulas obtained by different combinations of the two bands selected from the three bands are obtained, and then the degree of fit between the algae density index calculated by the different combination formulas and the measured algae density data is evaluated, and the combination formula with the best fit is confirmed as the algae density index quantification formula adapted to the target basin, and the algae density index value calculated under the algae density index quantification formula adapted to the target basin is determined, wherein: VI1=ρ1 / ρ2, VI2=(ρ1-ρ2) / (ρ1+ρ2), VI3=ρ1-ρ2, If the algae density index is greater than the threshold, the algae bloom situation exists; if the algae density index is less than the threshold, the algae bloom situation does not exist.

2. The method according to claim 1, characterized in that The method further comprises: The initial ground hyperspectral image is preprocessed, wherein the preprocessing includes radiation calibration, geometric correction, image mosaicking, spatial clipping, atmospheric correction, cloud removal, and band resampling.

3. The method according to claim 1, characterized in that The effective sample points are screened out based on the measured chlorophyll a content data of the ground measured points, including: The measured chlorophyll a content data is equally divided into multiple groups of chlorophyll a content data, each group of chlorophyll a content data intervals is cycled, the data points of chlorophyll a content and corresponding algae density data in each specified interval are screened out, and the data points in the specified interval with algae density data values ​​in the range of 5% to 95% are taken as the valid sample points.

4. The method according to claim 1, characterized in that: The algae density sample spectral library involves 9 bands, including an aerosol band of 443nm, a blue band of 490nm, a green band of 560nm, a red band of 665nm, a red band of 705nm, a red band of 740nm, a red band of 783nm, a red band of 843nm and a red band of 865nm.

5. A water bloom processing device based on ground hyperspectral images, characterized in that: The device comprises: An initial data acquisition unit is used to acquire, for a target watershed, the corresponding initial ground hyperspectral image, the measured algae density data of the ground measured point, the measured chlorophyll a content data of the ground measured point, and the initial geographical location data of the ground measured point; A spectral library construction unit is used to screen out valid sample points based on the measured chlorophyll a content data, determine target geographic location data of the valid sample points from the initial geographic location data, and then determine target spectral information of the target ground hyperspectral image corresponding to the valid sample points from the initial ground hyperspectral image according to the target geographic location data, and match the target spectral information with the measured algae density data to obtain an algae density sample spectral library; A model building unit, used to build a corresponding Gaussian process regression model based on the algae density sample spectrum library, wherein the input of the Gaussian process regression model is the spectrum data, and the output is the algae density data; A band determination unit, used to determine the algae density response capabilities of different spectral sensor bands based on the Gaussian process regression model, and determine three bands with the best algae density response capabilities; An index calculation unit, used to calculate the algae density index based on the three bands, and determine whether there is an algae bloom in combination with a corresponding algae density index threshold; A distribution map processing unit, used for processing geographical distribution maps of water bloom algae density indexes of different watersheds including the target watershed, and classifying them according to the size of the algae density index; The model building unit is specifically used for: Selecting an initialized Gaussian process regression model, taking the spectral information as an independent variable, taking the algae density data as a dependent variable, selecting a Gaussian process kernel function and a leave-one-out method as iterative optimization parameters, establishing the Gaussian process regression model, and recording it as f(x); Using the Monte Carlo method to simulate N sample points of d-type spectral information, an N×2d spectral information sample matrix is ​​generated, wherein each row of the N×2d spectral information sample matrix is ​​a sampling point in a 2-dimensional hyperspace; Using the first d columns of the N×2d spectral information sample matrix as spectral information matrix A and the remaining d columns as spectral information matrix B to give two independent samples of spectral information for N points in a d-dimensional unit hypercube; Constructing a dN×d spectral information matrix i=1,2,...,d, the dN×d spectral information matrix The i-th column of is the i-th column of the spectral information matrix B, and the remaining columns are the spectral information matrix A; The spectral information matrix A, the spectral information matrix B and the dN×d spectral information matrix N(d+2) points in the input space are in the spectral information matrix A, the spectral information matrix B and the dN×d spectral information matrix Run f(x) for each design point in , and give a total of N(d+2) values ​​of f(A), f(B), and value; The band determination unit is specifically used for: The spectral response index is calculated by the following formula: S Ti =E X~i (There is Xi (Y|X ~i )) / Yes(Y), Among them, E X~i (Var Xi (Y|X ~i )) means when X takes X ~i The conditional variance of the output when Y is represented by X ~i The algae density data obtained, X ~i refers to the i-th column of the input spectral information matrix, j represents the current sample point among the N sample points, Var(Y) is the variance of the algae density data, S Ti is a spectral response index of algae density data, wherein the spectral response index is proportional to the responsiveness of the algae density; The index calculation unit is specifically used for: The enumeration method is adopted to select two bands from the three bands and substitute them into the construction process of the algae density index quantification formula, wherein, in each construction process of the algae density index quantification formula, the two currently selected bands ρ1 and ρ2 are substituted into the quantification formulas of the ratio vegetation index VI1, the normalized vegetation index VI2, and the difference vegetation index VI3 for calculation, and then the combination formula consisting of the three elements of the ratio vegetation index VI1, the normalized vegetation index VI2 and the difference vegetation index VI3 is exhaustively enumerated in a constraint mode involving only addition, subtraction, multiplication and division operations, and different combination formulas obtained by different combinations of the two bands selected from the three bands are obtained, and then the degree of fit between the algae density index calculated by the different combination formulas and the measured algae density data is evaluated, and the combination formula with the best fit is confirmed as the algae density index quantification formula adapted to the target basin, and the algae density index value calculated under the algae density index quantification formula adapted to the target basin is determined, wherein: VI1=ρ1 / ρ2, VI2=(ρ1-ρ2) / (ρ1+ρ2), VI3=ρ1-ρ2, If the algae density index is greater than the threshold, the algae bloom situation exists; if the algae density index is less than the threshold, the algae bloom situation does not exist.

Citation Information

Patent Citations

  • Lake cyanobacterial bloom detection method

    CN117115077A

  • KR20230103612A